A new nipple position measure to enhance atlas-based segmentation in breast cancer radiotherapy planning

This technical report describes a method for template selection in atlas-based segmentation (ABS) for breast cancer that uses breast volume and nipple position.

Fifty patients who underwent postoperative radiotherapy (RT) after breast-conserving surgery were included. An experienced radiation oncologist manually segmented the bilateral breasts and recorded parameters including breast volume and nipple position. Nipple position was defined as the vertical distance from the midsternal skin to the nipple base on axial computed tomography. A radiation technologist subjectively selected 8 patients as templates and generated an ABS using each template for the remaining 49 cases (392 template-case pairs, 784 breast pairs). The Dice similarity coefficient (DSC) was used to assess the agreement between ABS and manual segmentation. Both breasts were evaluated simultaneously without consideration of the differences between the ipsilateral and contralateral operated sides. The Mann-Whitney U-test and Kruskal-Wallis test were used to compare DSC values among groups based on breast parameters.

Breast volume (P < .001) and nipple position (P < .001) were significant parameters for template selection in the 784 breast pairs. Selecting templates with breast volumes that were half to double those of the test case while maintaining the dorsal nipple position relative to the test case resulted in a median DSC of 0.88.

Breast volume and nipple position are valuable parameters that can improve template selection in ABS for breast cancer.

We developed a new nipple position measure to enhance ABS in breast cancer RT planning.